Nvidia disclosed $118 billion in maximum gross guarantee exposure and $32 billion in future additional neocloud commitments in its third-quarter fiscal 2027 CFO commentary, accompanying its November 17 results. The figures show how much financial support sits behind the AI infrastructure buildout, while the conditions explain when that support actually takes effect.
The company says its neocloud commitments depend on operators securing financing and deploying computing capacity that is operational and accepted. Commitments can be reduced or eliminated when those contingencies are not met. Its largest guarantee arrangement, up to $105 billion for OpenAI’s PORTS-Pike campus, activates in phases as individual buildings become ready for service. These are conditional commitments, not a statement that Nvidia has already paid out $118 billion. (Nvidia CFO commentary)
The distinction matters for infrastructure financing. A backstop available after a facility meets specified milestones provides different protection from one covering a project from the start of construction.
Guarantees Grew While Neocloud Commitments Fell
Nvidia’s two headline figures moved in opposite directions during the quarter. Future neocloud commitments fell by $4 billion, while maximum gross guarantee exposure increased by $9.5 billion, according to Blockspace’s account of the disclosure.
| Disclosed measure | Previous quarter | Q3 fiscal 2027 | Change |
|---|---|---|---|
| Future neocloud commitments | $36 billion | $32 billion | Down $4 billion |
| Maximum gross guarantee exposure | $108.5 billion | $118 billion | Up $9.5 billion |
That comparison complicates any simple claim that Nvidia is pulling back from financing the AI industry. One category contracted while another expanded. The disclosed support is changing in composition, and the categories have different contractual purposes and activation conditions.
Neoclouds are specialist cloud operators focused on AI computing. Nvidia’s commitments to those operators are reported separately from its guarantees, which provide financial backing under specified circumstances. Adding the two figures together and describing the result as current debt or cash spending would misrepresent the disclosure.
For scale, Nvidia reported quarterly revenue of $115.8 billion, including $107.6 billion from Data Center. Its maximum guarantee exposure is therefore slightly larger than one quarter’s revenue. That comparison illustrates size only: revenue measures activity during a quarter, while the guarantees cover potential obligations across future periods. (Nvidia CFO commentary)
Nvidia also said it had made no payments under the guarantees, according to Blockspace. That describes realized payments to date, not the likelihood or eventual size of future losses.
Financing and Deployment Control When Support Starts
Nvidia’s neocloud commitments require several steps before they take effect:
- Financing must be secured. The operator needs the capital required to carry out the project.
- Computing capacity must be deployed. Planned capacity alone does not satisfy the deployment condition.
- The capacity must be operational. Installed equipment must be ready to function.
- Nvidia must accept the capacity. Acceptance is another condition before the commitment becomes effective.
The company says commitments may be reduced or eliminated if the relevant contingencies are missed. This makes project execution central to understanding the numbers. An announced agreement can represent a substantial prospective commitment while remaining dependent on work and financing that have not yet been completed. (Nvidia CFO commentary)
Interim CEO and CFO Colette Kress also described conditions governing guarantees during the earnings call, according to TechCrunch’s November 18 report. Her explanation covered customers raising capital, preparing buildings, procuring computing equipment and installing it before the relevant guarantee begins.
Kress said this was a clarification of existing terms, not a policy change. That distinction is important: the disclosure provides a clearer public account of restrictions, but it should not be read as evidence that Nvidia has newly tightened every existing agreement.
The practical consequence is that securing financing remains a prerequisite for support under the disclosed conditions. A developer cannot treat the headline commitment as proof that Nvidia has already supplied the project’s construction capital.
The neocloud acceptance conditions and the building-readiness conditions for PORTS-Pike should also be kept distinct. They describe different arrangements, even though both make support contingent on milestones.
PORTS-Pike Accounts for About 89% of the Maximum
Nvidia’s largest disclosed arrangement is up to $105 billion of phased credit support for OpenAI’s PORTS-Pike campus. The guarantees activate as completed buildings become ready for service. (Nvidia CFO commentary)
Blockspace reports that the Ohio campus covers approximately 4.25 gigawatts across nine buildings and is leased to OpenAI for 20 years. The first guarantee is scheduled to activate in fiscal 2029. Its breakdown of Nvidia’s $118 billion maximum exposure is:
| Guarantee category | Maximum exposure |
|---|---|
| PORTS-Pike campus credit support | $105 billion |
| Other neocloud land, power and shell guarantees | $3.5 billion |
| Power and systems credit support | $9.5 billion |
| Total | $118 billion |
Source: Blockspace’s Q3 earnings coverage.
Dividing $105 billion by $118 billion puts PORTS-Pike at approximately 89% of the disclosed maximum. That concentration is consequential: the headline total is dominated by one campus arrangement, not spread evenly across a broad portfolio of unrelated sites.
I’d pay particular attention to that concentration alongside the activation terms. Building-by-building phasing ties support to delivered infrastructure, but the overall exposure remains heavily dependent on one development. Staging changes when exposure becomes effective; it does not make the underlying concentration disappear.
The Neocloud Schedule Extends Beyond Fiscal 2031
Nvidia’s $32 billion in future neocloud commitments is distributed across several fiscal years, with nothing scheduled for the remainder of fiscal 2027.
| Nvidia fiscal year | Future neocloud commitments |
|---|---|
| Remainder of 2027 | $0 |
| 2028 | $6 billion |
| 2029 | $7 billion |
| 2030 | $6 billion |
| 2031 | $5 billion |
| 2032 and thereafter | $8 billion |
| Total | $32 billion |
The schedule is reported in Blockspace’s analysis of the CFO commentary. These are Nvidia fiscal years, not calendar years, and the schedule remains subject to the disclosed contingencies.
Fiscal 2028 through fiscal 2030 account for $19 billion, or about 59% of the total. The final $8 billion is grouped into fiscal 2032 and later, so it should not be interpreted as one year’s spending.
This schedule describes a multi-year pipeline. It is separate from the PORTS-Pike guarantee timetable, whose first activation is scheduled for fiscal 2029. Treating both as immediate obligations would erase a meaningful difference in timing as well as contract structure.
Why Nvidia Is Focusing on Land, Power and Buildings
Kress said Nvidia would continue supporting important data-center projects while becoming more selective about the companies and arrangements it finances, TechCrunch reported.
Frequently Asked Questions
4 questions
1Has Nvidia Already Paid $118 Billion?
No. The figure is maximum gross guarantee exposure, not cash already paid. Nvidia said it had made no payments under the guarantees as of the disclosure.